XBP1 Signalling Is Essential for Alleviating Mutant Protein Aggregation in ER-Stress Related Skeletal Disease

RESEARCH ARTICLE XBP1 signalling is essential for alleviating mutant protein aggregation in ER-stress related skeletal disease

1 1 2¤ 1 Katarzyna A. PiroÂ gID *, Ella P. Dennis , Claire L. Hartley , Robert M. Jackson , 2 2 3 Jamie SoulID , Jean-Marc SchwartzID , John F. BatemanID , Raymond P. Boot- 2 1 HandfordID , Michael D. Briggs

1 Institute of Genetic Medicine, Newcastle University, Newcastle, United Kingdom, 2 Wellcome Trust Centre for Cell Matrix Research, University of Manchester, Manchester, United Kingdom, 3 Murdoch Children's a1111111111 Research Institute, Parkville, Victoria, Australia a1111111111 a1111111111 ¤ Current Address: Genomic Diagnostics Laboratory, Manchester Centre for Genomic Medicine, Manchester a1111111111 University Hospitals NHS Foundation Trust, Manchester, Manchester, United Kingdom a1111111111 * [email protected]

Abstract

OPEN ACCESS The unfolded protein response (UPR) is a conserved cellular response to the accumulation Citation: Piro´g KA, Dennis EP, Hartley CL, Jackson of proteinaceous material in endoplasmic reticulum (ER), active both in health and disease RM, Soul J, Schwartz J-M, et al. (2019) XBP1 to alleviate cellular stress and improve protein folding. Multiple epiphyseal dysplasia signalling is essential for alleviating mutant protein (EDM5) is a genetic skeletal condition and a classic example of an intracellular protein aggregation in ER-stress related skeletal disease. PLoS Genet 15(7): e1008215. https://doi.org/ aggregation disease, whereby mutant matrilin-3 forms large insoluble aggregates in the ER 10.1371/journal.pgen.1008215 lumen, resulting in a specific `disease signature' of increased expression of chaperones and Editor: Tom van Agtmael, University of Glasgow, foldases, and alternative splicing of the UPR effector XBP1. Matrilin-3 is expressed exclu- UNITED KINGDOM sively by chondrocytes thereby making EDM5 a perfect model system to study the role of Received: September 24, 2018 protein aggregation in disease. In order to dissect the role of XBP1 signalling in aggregation-related conditions we crossed a p.V194D Matn3 knock-in mouse model of EDM5 with a Accepted: May 27, 2019 mouse line carrying a cartilage specific deletion of XBP1 and analysed the resulting pheno- Published: July 1, 2019 type. Interestingly, the growth of mice carrying the Matn3 p.V194D mutation compounded Copyright: © 2019 Piro´g et al. This is an open with the cartilage specific deletion of XBP1 was severely retarded. Further phenotyping access article distributed under the terms of the revealed increased intracellular retention of amyloid-like aggregates of mutant matrilin-3 Creative Commons Attribution License, which permits unrestricted use, distribution, and coupled with dramatically decreased cell proliferation and increased apoptosis, suggesting reproduction in any medium, provided the original a role of XBP1 signalling in protein accumulation and/or degradation. Transcriptomic analy- author and source are credited. sis of chondrocytes extracted from wild type, EDM5, Xbp1-null and compound mutant lines Data Availability Statement: The full datasets are revealed that the alternative splicing of Xbp1 is crucial in modulating levels of protein aggre- available from the NCBI Gene Expression Omnibus gation. Moreover, through detailed transcriptomic comparison with a model of metaphyseal (GEO), accession number GSE120308 (http:// www.ncbi.nlm.nih.gov/geo/). chondrodysplasia type Schmid (MCDS), an UPR-related skeletal condition in which XBP1 was removed without overt consequences, we show for the first time that the differentiation- Funding: This work was funded by grants from the Wellcome Trust (https://wellcome.ac.uk/ grant state of cells within the cartilage growth plate influences the UPR resulting from retention of 084353/Z/07/; MDB) and EU (SYBIL - European a misfolded mutant protein and postulate that modulation of XBP1 signalling pathway pres- Community’s Seventh Framework Programme ents a therapeutic target for aggregation related conditions in cells undergoing proliferation. (http://ec.europa.eu/research/fp7/index_en.cfm, FP7/2007-2013; MDB, KAP, JMS, RBH) under grant agreement no.602300). The funders had no

PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008215 July 1, 2019 1 / 30 UPR is an essential component of genetic skeletal disease

role in study design, collection, analysis or interpretation of data, or writing the manuscript, or decision to submit the manuscript. Author summary

Competing interests: The authors have declared A significant proportion of genetic skeletal diseases result from misfolding of structural pro- that no competing interests exist. teins within the endoplasmic reticulum (ER) of cartilage cells (chondrocytes). Interestingly, these diseases often share transcriptomic changes with non-skeletal conditions such as Alzhei- mer’s disease and diabetes and include changes in the unfolded protein response (UPR). UPR signals are conveyed through three effector molecules, IRE1, ATF6 and PERK, understanding of which is of utmost importance for the development of novel therapeutic approaches. IRE1 is the most conserved UPR pathway, with XBP1 as its main effector. Interestingly, in our study we show that chondrocyte response to ER stress is highly modulated by the differentiation state of the cell. We have deleted the IRE1/XBP1 pathway activity in the chondrocytes of a mouse model of multiple epiphyseal dysplasia (MED), expressing misfolding matrilin-3 predominantly in proliferating chondrocytes, and in metaphyseal chondrodysplasia type Schmid (MCDS), resulting from mutations in type X collagen, expressed by the hypertrophic chondrocytes. Here, using deep phenotyping and extensive transcriptomic analysis we demonstrate that whilst the IRE1/XBP1 pathway is redundant in hypertrophic chondrocyte pathology, it is in fact essential for UPR responses in proliferating cells, making this highly conserved pathway an attractive therapeutic target for a broad range of UPR related conditions.

Introduction The unfolded protein response (UPR) is one of the canonical cellular stress pathways that is triggered by unfolded proteins accumulating in the ER lumen. The pathway is active in both health and disease and many secretory cells have a highly active UPR to allow a greater secretory output [1]. Over the recent years, the UPR triggered by ER-stress has been recognised as a crucial component in the pathobiology of many human diseases, including neurodegenerative conditions, diabetes and numerous musculoskeletal phenotypes [2–6]. However, the specific role of the UPR in the context of human disease is still being determined and it is hoped that it will offer attractive therapeutic targets and avenues in the future. The canonical UPR response is initiated when the chaperone protein BiP dissociates from its three membrane bound receptors, PERK, ATF6 and IRE1 [7]. This dissociation is triggered by the exposed hydrophobic residues of misfolded proteins in the ER lumen and the UPR then pro- ceeds along the three signalling pathways, which are further modulated by the levels and duration of the stress [8]. Furthermore, these branches cross talk and signal in conjunction with, or modulate other important mechanisms such as inflammation and autophagy, thereby offering the cell a robust machinery to counteract protein misfolding and oxidative stress. Following dissociation of BiP, PERK dimerises and autophosphorylates, which in turn triggers the phosphorylation of elongation factor eIF2ɑ. This leads to attenuation of general protein translation allowing the cell to recover from an abnormal protein load [9]. However, several proteins escape this translational block, including ATF4, a downstream effector of PERK, which can trigger ER-stress related apoptosis via CHOP (or DDIT3 [10, 11]. Following release from BiP, ATF6 translocates to the Golgi apparatus where it is cleaved, releasing an active transcription factor. ATF6 signalling then leads to an upregulation of chaperones and XBP1 expression, but it can also trigger apoptosis via CHOP mediated signalling [12]. The third transmembrane sensor is IRE1, which dimerises and autophosphorylates upon the dissociation of BiP. In the active form IRE1 can induce the alternative splicing of XBP1, producing an active transcription factor which upregulates chaperone genes and genes responsible for ER-associated degradation of accumulated proteins (ERAD) [13–15]. It is therefore not surprising that the XBP1 branch of the UPR pathway has been

PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008215 July 1, 2019 2 / 30 UPR is an essential component of genetic skeletal disease

implicated in many protein aggregation diseases including Alzheimers [16, 17], Huntington’s disease [18, 19], type II diabetes [20] and several skeletal conditions [5, 7, 12, 21] amongst others. Multiple epiphyseal dysplasia (MED) is predominantly an autosomal dominant skeletal dysplasia characterised by disproportionate short-limbed dwarfism and early onset joint degenera- tion [22]. MED results from dominant-negative mutations in three structural proteins of the cartilage extracellular matrix (ECM); cartilage oligomeric matrix protein (COMP), matrilin-3 and type IX collagen, which interact with each other in the ECM [23]. Matrilin-3 is a tetrameric bridging molecule that regulates collagen fibrillogenesis [24] and each monomer consists of a single von Willebrand factor A like domain (A-domain), four EGF-like repeats and an oligomer- isation domain. MED-causing mutations (EDM5; OMIM #607078) are located exclusively in the disulphide bond stabilised A-domain and result in misfolding and retention of the mutant protein in the ER lumen. A classical UPR is activated with an upregulation of generic chaperones as well as a more specific cocktail of disulphide isomerases such as CRELD2, PDIA1, PDIA3 and PDIA6 [5]. However, the mutant protein is prone to aggregation, forming large insoluble non- native disulphide bonded aggregates in the ER that contain a high percentage of ß-sheet folds [5, 25], suggesting a propensity to form amyloid-like deposits upon misfolding [26] that appear resistant to degradation. This in turns leads to the dysregulation of chondrocyte apoptosis, a decrease in chondrocyte proliferation and consequently reduced bone growth [27, 28]. We have previously demonstrated that chondrocytes from a mouse model of EDM5 with a p.V194D mutation in Matn3 exhibited a specific upregulation of genes in the XBP1 branch of the UPR [5, 29, 30]. A similar upregulation of XBP1 signalling was seen in the Col10a1 N617K model of metaphyseal chondrodysplasia type Schmid (MCDS), but not in an allelic series of pseudoachondroplasia (PSACH) causing Comp mutations, indicating gene product specificity of this arm of the UPR [6, 31, 32]. Therefore, in order to further understand the role of XBP1 signalling in the protein aggregation and disease pathology of Matn3-related MED we crossed a p.V194D knock-in mouse model of EDM5 with a mouse line carrying a cartilage specific deletion of Xbp1 and analysed in-depth the resulting phenotype.

Results Engineering the chondrocyte specific knock out of Xbp1 in the EDM5 mouse model We have previously generated a mouse model of EDM5 (p.V194D in Matn3 and referred to as Xbp1WT Matn3V194D in this paper, [5]). Interestingly, RT-PCR and sequencing of cDNA derived from wild type (Xbp1WT Matn3WT) and homozygous mutant (Xbp1WT Matn3V194D) cartilage dissected from 5-day-old mice revealed non-conventional splicing of Xbp1 in Xbp1WT Matn3V194D chondrocytes (Fig 1A). The Xbp1WT Matn3V194D mouse line was therefore crossed with a mouse line in which Xbp1 had been rendered inactive in chondrocytes through the Col2a1-Cre/loxP-mediated deletion of exon 2 (Xbp1Col2CreΔex2, [21, 31]), in order to study the role of XBP1 signalling in EDM5. This breeding strategy generated the Xbp1Col2CreΔex2 Matn3V194D mouse line. Xbp1Col2CreΔex2 Matn3V194D mice were viable and fertile; however, mice homozygous for both mutant alleles had breeding and survival complications due to their dramatically reduced size, breathing difficulties and narrower birth canals.

The loss of XBP1 from cartilage leads to an exacerbation of the EDM5 phenotype Bone measurements were used to determine the effect of XBP1 deletion on endochondral (tibia and femur lengths) and intramembranous (inner canthal distance) ossification of the

PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008215 July 1, 2019 3 / 30 UPR is an essential component of genetic skeletal disease

Fig 1. (A) Mutation in MATN3 induced an UPR response and alternative splicing of Xbp1. (B) Xbp1Col2CreΔex2 mice presented with a mild chondrodysplasia and were shorter than their wild type littermates. Xbp1WTMatn3V194D mice presented with typical EDM5 short limbed dwarfism. The skeletal phenotype of Xbp1Col2CreΔex2 Matn3V194D was very severe, with short limbs, constricted ribcages and a pronounced hip dysplasia. (C) Bone measurements using Xray analysis at 3, 6 and 9 weeks of age show disruption of endochondral ossification (tibia, femur), but not the intramembranous ossification (inner canthal distance, ICD) in the Xbp1Col2CreΔex2, Xbp1WT Matn3V194D and Xbp1Col2CreΔex2 Matn3V194D mice (n�5, Student t-test). Key: Xbp1u - unspliced Xbp1, Xbp1s –spliced Xbp1, �� P<0.001. https://doi.org/10.1371/journal.pgen.1008215.g001

Xbp1WT Matn3V194D mouse model. The Xbp1Col2CreΔex2 mice were slightly shorter than their wild type littermates, as previously reported, indicating a role for Xbp1 in normal skeletal development [21]. Xbp1WT Matn3V194D mice were shorter than both wild type mice and

PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008215 July 1, 2019 4 / 30 UPR is an essential component of genetic skeletal disease

Xbp1Col2CreΔex2 mice with a comparable genetic background [33]. Unsurprisingly, the Xbp1Col2CreΔex2 Matn3V194D mice had a more pronounced short-limbed dwarfism than that previously reported for the Xbp1WT Matn3V194D mice, signifying a crucial protective role of the XBP1 branch of UPR in EDM5 pathology (Fig 1B and 1C). Paradoxically, this is in direct contrast to the minor role for XBP1 signalling proposed in the recent study of the Col10aN617K model of metaphyseal chondrodysplasia type Schmid (MCDS) [31]. The Xbp1Col2CreΔex2 Matn3V194D mice had dramatically shorter long bones (>30% reduction compared to wild type mice and ~20% reduction compared to the Xbp1WT Matn3V194D mice) and abnormal bell-shaped rib cages, which appeared to hinder their ability to breathe correctly. The inner canthal distance (ICD) was not altered in any of the mice studied indicating that intramembranous ossification was not affected. Over time there was severe truncation and rotation of the limb, abnormal bending of the long bones and severe constriction of the rib cages in Xbp1Col2CreΔex2 Matn3V194D mice (S1 Fig).

Ablation of XBP1 dramatically altered Matn3V194D cartilage growth plate morphology Deletion of XBP1 in the Xbp1WT Matn3V194D mouse line severely affected the morphology of the cartilage growth plates (Fig 2). Briefly, the Xbp1WT and Xbp1Col2CreΔex2 growth plates presented with a typical and well-organised columnar arrangement of chondrocytes in the proliferative zone and an ordered progression from the resting to proliferative to hypertrophic cells along the vertical axis of the growth plate. The growth plates from Xbp1Col2CreΔex2 mice had a slightly reduced hypertrophic zone and small areas of hypocellularity, consistent with the previously published study [21]. In contrast, growth plates from Matn3V194D mice were characterised by enlarged cells in the resting and proliferative zones due to retention of misfolded mutant matrilin-3 as previously described [33]. Finally, the growth plates from Xbp1Col2CreΔex2 Matn3V194D mice had a dramatically altered morphology with abnormally enlarged cells present throughout the entire growth plate and concurrent with an apparent increase in the retention of mutant matrilin-3. This increased retention of mutant matrilin-3 appeared to correlate with an increase in amyloid-like intracellular deposits detected by Congo Red fluorescence (S2 Fig). The severe disorganisation of the growth plates from the Xbp1Col2CreΔex2 Matn3V194D mice rendered impractical any measurement of the respective zones using histology images. However, the distribution of type X collagen (a marker of chondrocyte hypertrophy) in the Xbp1Col2CreΔex2 Matn3V194D growth plates at 3 weeks was altered and collagen X staining extended into the proliferative zone, coinciding with the abnormally enlarged chondrocytes and indicating accelerated differentiation. In contrast, staining for type II collagen (a major component of the cartilage ECM) was unaffected (Fig 2).

Chondrocyte apoptosis and proliferation are affected by the deletion of XBP1 in Matn3V194D mice Chondrocyte proliferation in the growth plates of both Xbp1WT Matn3V194D and Xbp1Col2- CreΔex2 mice was significantly reduced when compared to wild type controls as previously reported (~16% and ~40% decrease respectively) [21, 33]. By comparison, chondrocyte proliferation in Xbp1Col2CreΔex2 Matn3V194D mice was decreased by ~80% when compared to the wild type mice and by ~66% when compared to Xbp1WT Matn3V194D mice suggesting a synergistic effect of the two mutations (Fig 3A). Moreover, the staining for BrdU not only showed a decrease in the relative number of BrdU positive cells, but also a reduced intensity of staining of these positive cells, suggesting a slower rate of DNA synthesis and an impaired cell cycle (Fig 3B).

PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008215 July 1, 2019 5 / 30 UPR is an essential component of genetic skeletal disease

Fig 2. Immunohistochemical analysis of the cartilage growth plates at 3 weeks showed no difference in matrilin-3 (brown staining) amount or the typical extracellular distribution in the Xbp1Col2CreΔex2 tissue. The Xbp1 null growth plates showed small areas of hypocellularity consistent with the previously reported decrease in proliferation (box). Mutant matrilin-3 was intracellularly retained (arrowhead) and depleted from the extracellular matrix in Xbp1WT Matn3V194D cartilage. Moreover, the columnar organisation of the proliferative one was affected and the cells adopted a more rounded morphology. The removal of XBP1 from EDM5 cartilage (Xbp1Col2CreΔex2 Matn3V194D) resulted in a dramatically aggravated phenotype with increased

PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008215 July 1, 2019 6 / 30 UPR is an essential component of genetic skeletal disease

intracellular retention of matrilin-3 and the appearance of abnormally enlarged cells throughout the tissue (oval). Nuclear fast green was used as a counterstain. The zone of type X collagen (red) staining was broader in the Xbp1 Col2CreΔex2Matn3V194D growth plates at 3 weeks, reflecting the growth plate disorganisation. Type II collagen was not affected in any of the analysed mouse models. DAPI was used as nuclear counterstain (blue).Key: RZ–resting zone, PZ–proliferative zone, HZ–hypertrophic zone, scale bar 100μm. https://doi.org/10.1371/journal.pgen.1008215.g002

Chondrocyte apoptosis in the growth plate was not affected by the deletion of XBP1 from cartilage as previously reported (Xbp1Col2CreΔex2 compared to Xbp1WT controls)[21]. In contrast, the p.V194D Matn3 mutation had a slight negative effect on the levels of apoptosis in the hypertrophic and proliferative zones, which is consistent with previous observations [33]. However, in the Xbp1Col2CreΔex2 Matn3V194D double mutant mice there was a dramatic increase in the relative levels of chondrocyte apoptosis in the resting, proliferative and hypertrophic zones (1.5, ~8 and ~8-fold respectively compared to the wild type controls; Fig 3C). Indeed, TUNEL positive cells were found throughout the growth plates of Xbp1Col2CreΔex2 Matn3V194D mice and appeared to correlate with the enlarged cell morphology previously noted in the histological analysis (Fig 3D).

Comparative microarray analysis confirms that the XBP1 branch of the UPR has a role in modulating the aggregation of mutant matrilin-3 in the chondrocytes of Matn3V194D mice Deleting XBP1 from the chondrocytes of Xbp1WT Matn3V194D cartilage further exacerbated the disease phenotype, indicating that the XBP1 branch of the UPR pathway had a ‘chondro- protective’ role in proliferating chondrocytes. To further define the putative protective role of XBP1 we performed microarray analyses on mRNA derived from wild type and mutant ani- mals with a comparable C57BL6 genetic background. Volcano plots showing the differential expression of genes and significant changes are shown in Fig 4A–4D and a heat map comparison is shown in S3 Fig. In total 2092 genes were differentially expressed in mutant Matn3 chondrocytes (Xbp1WT Matn3V194D vs Xbp1WT analysis) compared to 2396 in XBP1 null cartilage (Xbp1Col2CreΔex2 vs Xbp1WT), 1316 in Xbp1Col2CreΔex2 Matn3V194D vs Xbp1Col2CreΔex2 and finally 712 in Xbp1Col2CreΔex2 Matn3V194D vs Xbp1WT Matn3V194D (S3 Fig). Overall, this indicated that both the presence of the matrilin-3 mutation and the absence of XBP1 were strong effectors of cartilage homeostasis. The highest number of differentially expressed genes in common (1192) was noted for the cartilage-specific deletion of XBP1 (Xbp1Col2CreΔex2 vs Xbp1WT) and the p.V194D Matn3 mutant (Xbp1WT Matn3V194D vs Xbp1WT) comparisons, suggesting that genes downstream of the XBP1 signalling pathway are key disease modulators in matrilin-3 related MED. Xbp1WT Matn3V194D vs Xbp1WT analysis confirmed the findings previously reported for this mouse line [5]. The 1494 genes downregulated in the mutant Matn3 chondrocytes (Xbp1WT Matn3V194D vs Xbp1WT) were predominantly associated with regulation of gene expression, cell proliferation and apoptosis. A distinct subset of downregulated genes was associated with modulating the aggregation of mutant/misfolded proteins (Cryab, Hspa1l, Hspb8, DNaja1). Genes associated with cartilage and bone development (such as Alpl, Col1a1, Col10a1, Evc, Hif1a, Igf1, Igfbp5, Ibsp, Mmp14, Nog, Pthlh, Smo and Tnc) were also decreased. In contrast, 598 genes significantly upregulated in the Matn3 mutant chondrocytes (Xbp1WT Matn3V194D vs Xbp1WT) were predominately associated with cell proliferation, migration and the cellular response to ER stress [5] and included genes such as Creld2, Canx, Dnajc3, Grp94, Hyou1, Manf (Armet), Pdia3, Pdia4, Trib3 and Xbp1 (S1 Table).

PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008215 July 1, 2019 7 / 30 UPR is an essential component of genetic skeletal disease

PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008215 July 1, 2019 8 / 30 UPR is an essential component of genetic skeletal disease

Fig 3. (A) 2h BrdU labelling pulse was performed to assess cell proliferation in the growth plate cartilage at 3 weeks of age. Xbp1 null cartilage showed a 16% decrease in proliferation compared to the wild type control. The proliferation in the Xbp1WT Matn3V194D cartilage was decreased by nearly 40%. Proliferation levels in the Xbp1Col2CreΔex2 Matn3V194D cartilage reflected the synergistic effect of the two genetic insults and was decreased by 66% (n = 3, One Way ANOVA). (B) The incorporation of the labelling agent (purple staining) into the proliferating cells was also decreased in the Xbp1Col2CreΔex2 Matn3V194D chondrocytes indicating a slowed down cell cycle. Nuclear Fast Green was used as counterstain. (C) TUNEL assay was performed to assess apoptosis in cartilage growth plates at 3 weeks (n = 3, One Way ANOVA). Apoptosis was dramatically increased in all zones of the Xbp1Col2CreΔex2 Matn3V194D growth plate. (D) The apoptosis, normally occurring at the lower hypertrophic zone in the cartilage growth plate, was also dysregulated in the Xbp1Col2CreΔex2 Matn3V194D with cells dying in the proliferative and resting zones as well. Interestingly, the cells undergoing apoptosis correlated with the abnormally enlarged cellular morphology. Positive cells shown in green, DAPI used as nuclear counterstain (blue). Key: RZ–resting zone, PZ–proliferative zone, HZ–hypertrophic zone, � P<0.05, �� P<0.01, �� P<0.001, scale bar 100μm. https://doi.org/10.1371/journal.pgen.1008215.g003

A Xbp1Col2CreΔex2 vs Xbp1WT comparison confirmed the previously published involvement of Xbp1 in bone formation [21] and modulation of the ER stress response [34]. 1114 genes downregulated in XBP1 null cartilage (Xbp1Col2CreΔex2 vs Xbp1WT) were associated with transcription regulation, apoptosis, protein ubiquitination and cell cycle regulation [21]. Several genes involved in protein folding were also decreased including Cryab, Dnajc4, Dnajc13, Dnajc22, Hspa4l and Hspa8. 1282 genes upregulated in Xbp1 null cartilage pertained to regulation of cell migration, angiogenesis and extracellular matrix organisation as previously published [21]. Genes for several ECM proteins including Col1a1, Col10a1 Col4a1, Col4a2, Col18a1, Fbln5 and Ibsp were also upregulated, together with several signalling molecules such as Fgf18, Fgf2, Notch1, Tgfb2 and Vegfc, indicating a deregulation of chondrocyte differentiation in the Xbp1Col2CreΔex2 mice. A comparative analysis of genes differentially regulated by the Matn3 mutation (Xbp1WT Matn3V194D vs Xbp1WT) with the genes differentially regulated by Xbp1 ablation in Matn3 mutant chondrocytes (Xbp1Col2CreΔex2 Matn3V194D vs Xbp1WT Matn3V194D) was undertaken to explain the dramatically exacerbated disease phenotype in the Xbp1Col2CreΔex2 Matn3V194D mice and to identify Xbp1-dependent pathways in EDM5 (Fig 4E and 4F). Interestingly, the differentially changed genes are involved in cartilage differentiation/dedifferentiation pathways (“lipolysis in adipocytes”, “mineral deposition”, “TGFß signalling”, “rheumatoid arthri- tis”) and in the UPR (“protein processing in the endoplasmic reticulum”; Fig 4F). A detailed analysis revealed 57 genes decreased in the Matn3 mutant chondrocytes and further decreased upon XBP1 deletion, indicating their expression is modulated downstream of XBP1. The GO terms for these genes included “protein folding”, “phospholipid biosynthesis process” and “nucleosome assembly” (S2 Table). The most highly and significantly represented GO term was “protein folding” and included the chaperone molecule DNAJA4, heat shock protein HSPA8 and crystallin alpha B (CRYAB) (S2 Table); all of which lie downstream of XBP1 signalling [34]. Moreover, the genes associated with “phospholipid biosynthesis process” included choline kinase alpha (Chka), ethanolaminephosphotransferase 1 (Ept1) and phosphatidylser- ine decarboxylase pseudogene 3 (Pisd-ps3), involved in the maintenance of vesicular membranes and in protein folding respectively. 150 genes were upregulated in the Matn3 mutant cartilage (Xbp1WT Matn3V194D vs Xbp1WT) and downregulated upon removal of XBP1 (Xbp1Col2CreΔex2 Matn3V194D vs Xbp1WT Matn3V194D) indicating an ER-stress triggered XBP1-dependent response (S3 Table). The top GO terms associated with these were “response to toxic substance”, “response to lipopolysac- caride”, “regulation of ERK1 and ERK2 cascade” and “cell migration”. The genes changed in the “ERK1 and ERK2 signalling pathway” and “cell migration” (Abl2, Cd44, Ccl5, Prkca, Sema7a) are known to respond to extracellular signals and may be reflecting extracellular changes resulting from the intracellular stress. Ltbp1, one of the major players in the TGFß signalling, was dramatically increased in Matn3 mutant cartilage (Xbp1WT Matn3V194D vs Xbp1WT) and decreased upon Xbp1 removal, but unchanged in the XBP1 null control, indicating a stress-related response. Other interesting genes upregulated in Matn3 mutant cartilage

PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008215 July 1, 2019 9 / 30 UPR is an essential component of genetic skeletal disease

PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008215 July 1, 2019 10 / 30 UPR is an essential component of genetic skeletal disease

Fig 4. (A-D) Volcano plots of differential gene expression (in black) highlighting the significantly changed genes (1.5 fold, P<0.05; in red). (E) A Venn diagram comparing the significantly differentially expressed genes in the original EDM5 (Xbp1WT Matn3V194D vs Xbp1WT Matn3WT) model and the effect of XBP1 deletion (Xbp1Col2CreΔex2 Matn3V194D vs Xbp1WT Matn3V194D). (F) Pathway analysis of the Xbp1Col2CreΔex2 Matn3V194D vs Xbp1WT Matn3V194D microarray data was performed using the iPathwayGuide platform (Advaita Corp). Differentially expressed KEGG pathways are plotted using the P-values and the number of differentially expressed probes. https://doi.org/10.1371/journal.pgen.1008215.g004

but downregulated upon removal of Xbp1 were Mmp3, Mmp10, Adam19 and Sost, suggesting potential changes in chondrocyte maturation and differentiation. Moreover, several genes upregulated in mutant Matn3 cartilage and decreased upon removal of Xbp1 (Abcb1a, Creld2, Pdia6, Dnajc3, Ero1lb, Hao1, Hyou1, Magt1, Pex11a, Sdf2l1, Ugt1a1) reflected changes in protein folding, disulphide bond formation, peroxisome activity and drug metabolism machinery implying an Xbp1-dependent non-canonical stress pathway is activated by the accumulation of misfolded aggregated matrilin-3. Interestingly, some of these genes were also implicated in the pathobiology of other aggregation-related diseases such as Alzheimer’s [17, 35]. A total of 79 genes were downregulated in Matn3 mutant chondrocytes (Xbp1WT Matn3V194D vs Xbp1WT analysis) but upregulated in Matn3 mutant chondrocytes lacking XBP1 (Xbp1Col2CreΔex2 Matn3V194D vs Xbp1WT Matn3V194D) and the main GO term in this comparison was “osteoblast differentiation” (S4 Table). The genes associated with this group- ing (Col1a1, Igf1, Igfbp5, Ibsp, Tnc) were downregulated by the presence of matrilin-3 mutation on both Xbp1 backgrounds (wild type and null), indicating their downregulation in response to aggregation stress, and upregulated by the absence of XBP1 on wild type or matrilin-3 mutant background thereby suggesting XBP1 regulation [21]. Only 4 genes were increased in both Xbp1WT Matn3V194D vs Xbp1WT and in Xbp1Col2CreΔex2 Matn3V194D vs Xbp1WT Matn3V194D and these were Serpina3c, Serpina3n, Ptger and a hypo- thetical protein (2310033P09Rik).

Differential expression profiling helps separate the aggregation-dependent and the XBP1-dependent molecular events Expression profiling of the Matn3 mutant cartilage (Xbp1WT Matn3V194D vs Xbp1WT) compared to the Matn3 mutant cartilage on the XBP1-null background (Xbp1Col2CreΔex2 Matn3V194D vs Xbp1Col2CreΔex2) was performed in order to elucidate the molecular events specifically dependent upon matrilin-3 mutation and independent of the UPR signalling downstream of XBP1 (Figs 5 and 6). 47 genes were upregulated (Fig 5B) and 246 genes downregulated (Fig 6B) in the presence of the matrilin-3 mutation. The upregulated genes clustered in “metabolic process” and “protein folding” GO terms and included Bcat2, Lpcat1, Pnpla3 and Uap1, and Atf5, Canx, Derl3, Manf, Pdia3, Pdia4 and Pdia6 respectively. The downregulated genes pertained to “endochondral ossification” and included Alpl, Col1a1, Col10a1, Cyr61, Dlx5, Gabbr1, Igf1, Ibsp, Mef2c, Mmp14, Pthlh and Tnc; and “protein folding” including Cryab, Dnaja1, Dnaja4, Hspa4l, Hspa8, Hsp90aa1, Tubb5. Interestingly, a subset of ER-stress related genes downregulated independently of XBP1 in the presence of mutant matrilin-3 (Cryab, Dnaja4, Hspa8 and Hsp90aa1) were further decreased in the Xbp1Col2CreΔex2 Matn3V194D vs Xbp1WT Matn3V194D analysis and decreased in the Xbp1 null cartilage (Xbp1Col2CreΔex2 vs Xbp1WT analysis), indicating a synergistic effect of the two stressors and potential XBP1 modulation. 62 genes were upregulated (Fig 5C) and 165 genes downregulated (Fig 6C) following the removal of Xbp1 from chondrocytes with or without the Matn3 mutation (i.e. Xbp1Col2CreΔex2 vs Xbp1WT compared to Xbp1Col2CreΔex2 Matn3V194D vs Xbp1WT Matn3V194D) indicating UPR- independent XBP1 signalling. The upregulated genes largely pertained to extracellular matrix

PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008215 July 1, 2019 11 / 30 UPR is an essential component of genetic skeletal disease

PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008215 July 1, 2019 12 / 30 UPR is an essential component of genetic skeletal disease

Fig 5. (A) Venn diagram showing an overlap in the significantly upregulated microarray probes from the 4 analyses (Xbp1WT Matn3V194D vs Xbp1WT, Xbp1Col2CreΔex2 Matn3V194D vs Xbp1Col2CreΔex2, Xbp1Col2CreΔex2 Matn3V194D vs Xbp1WT Matn3V194D and Xbp1Col2CreΔex2 vs Xbp1WT. (B) A graph showing the GO terms associated with 47 upregulated probes representing the XBP1 independent effect of a MATN3 mutation. (C) A graph showing the GO terms associated with 62 upregulated probes representing the MATN3 mutation independent effect of XBP1 deletion. https://doi.org/10.1371/journal.pgen.1008215.g005

organisation and osteoblast differentiation (e.g. Col1a1, Igf1, Igfbp3, Igfbp5, Ibsp, Vegfc), which is in agreement with previously published data [21], whereas the downregulated genes were clustered to RNA splicing (e.g. Prpf38b, Rbm25, Rbm5, Cpsf6, Hspa8, Mbnl1, Srrm2, Tra2a) and protein folding (Ahsa1, Pdia6).

Aggregation-specific ER-resident chaperones are differentially expressed in chondrocytes from the MCDS and EDM5 mice, delineating a differentiation state specific stress response Removal of Xbp1 had a profound effect on the phenotype of EDM5 mice, which is a disease predominantly affecting chondrocytes undergoing proliferation, but in contrast, deletion of Xbp1 had no effect on the severity of MCDS, a disease of the hypertrophic zone of the growth plate [31]. We therefore assessed the modulation of the UPR machinery in different zones by comparing entire growth plate microarrays (Xbp1Col2CreΔex2 vs Xbp1WT) against a dataset generated for the hypertrophic zone only (Xbp1Col2CreΔex2(HZ) vs Xbp1WT(HZ); GEO series accession number GSE72261). Interestingly, both the hypertrophic zone alone and the full cartilage growth plate showed unique gene expression signatures upon removal of Xbp1 (S3 Fig). For example, whilst Pdia6 was decreased in both the hypertrophic zone and the full growth plate following XBP1 ablation, the genes encoding other aggregation-specific chaperones such as Cryab, Dnaja4, Hspa1l and Hspa8 were decreased in chondrocytes from the whole growth plate, but not in hypertrophic zone chondrocytes alone, suggesting differentiation-state dependent XBP1 modulation. In contrast, ATF6 signalling appeared to be affected by the deletion of XBP1 in both the whole growth plate and the hypertrophic zone alone. In the hypertrophic zone this resulted in an upregulation of the more stable, but weaker activator ATF6ß, whereas the full growth plate analysis showed an increase in ATF6ɑ and no change in ATF6ß, potentially indicating a decrease in ATF6ß in the proliferative zone and a differentiation-state dependent modulation of ATF6 signalling. We then compared the microarray data generated using hypertrophic chondrocytes from the MCDS mice [36] with the microarray data of chondrocytes from the EDM5 mice to determine the cellular response to the retention and aggregation of mutant protein (S3 Fig). 586 differentially expressed genes were shared between the MCDS (Col10a1N617K vs. Col10a1WT) and EDM5 (Matn3V194D vs. Matn3WT) cartilage. 136 genes were upregulated in both mice and the main GO term for these was “ER response” and included the following genes, Atf5, Atf6, Creld2, Derl3, Dnajc3, Hsp90b1 (Grp94), Hyou1, Manf, Pdia3, Pdia4, Trib3 and Xbp1. Of these genes Creld2, Dnajc3 and Hyou1 appeared to be controlled by XBP1 under ER stress as suggested by their downregulation in the Xbp1Col2Δex2 Matn3V194D vs. Xbp1WT Matn3V194D comparison. In contrast, Atf5, Derl3, Hsp90b1, Manf and Pdia4 were all upregulated or unchanged in Xbp1Col2Δex2 Matn3V194D vs. Xbp1Col2Δex2 and in Xbp1WT Matn3V194D vs. Xbp1WT comparisons, indicating an XBP1-independent effect of mutant protein misfolding/aggregation. Inter- estingly, several of these genes have previously been shown to be downstream of ATF6 signalling ([34]; Table 1). 169 genes were downregulated in both comparisons and the top GO term for these was “skeletal development”, including genes such as Alpl, Bcan, Cebpd, Col1a1, Crlf3, Deaf1, Foxa2, Ibsp, Kazald1, Mmp14, Nog, Smo and Sox4. Many of these represent a

PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008215 July 1, 2019 13 / 30 UPR is an essential component of genetic skeletal disease

PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008215 July 1, 2019 14 / 30 UPR is an essential component of genetic skeletal disease

Fig 6. (A) Venn diagram showing an overlap in the significantly downregulated microarray probes from the 4 analyses (Xbp1WT Matn3V194D vs Xbp1WT, Xbp1Col2CreΔex2 Matn3V194D vs Xbp1Col2CreΔex2, Xbp1Col2CreΔex2 Matn3V194D vs Xbp1WT Matn3V194D and Xbp1Col2CreΔex2 vs Xbp1WT. (B) A graph showing the GO terms associated with 47 downregulated probes representing the XBP1 independent effect of a MATN3 mutation. (C) A graph showing the GO terms associated with 62 downregulated probes representing the MATN3 mutation independent effect of XBP1 deletion. https://doi.org/10.1371/journal.pgen.1008215.g006

delay in terminal differentiation, which could explain the phenotype of short-limbed dwarfism in both mouse models. ER stress related genes that were differentially expressed between the two models included Ahsa1, Cryab, Hspa8, Hsp90aa1 and Hsph1, all of which were downregulated in Matn3 mutant (Matn3V194D vs. Matn3WT) cartilage and upregulated in the Col10a1

Table 1. A table summarising microarray data of the differential expression of representative IRE1ɑ (XBP1), ATF6 and PERK target genes [34, 37–40] in the EDM5 and MCDS models, and in EDM5 upon removal of XBP1 (EDM5/X). n.s. = not significant. Xbp1 target genes EDM5 EDM5/X MCDS Abcb1a 4.57 -3.19 n.s. Edem3 n.s. -2.28 2.64 Hao1 16.67 -2.60 n.s. Pdia6 1.60 -1.86 4.93 Pex11a 1.62 -1.99 n.s. Ugt1a1 1.90 -3.47 n.s. ATF6 target genes EDM5 EDM5/X MCDS Canx 1.86 n.s. 1.87 Derl3 25.20 n.s. 19.7 Ero1l 1.86 n.s. 8.57 PERK target genes EDM5 EDM5/X MCDS Ddit3 -2.28 n.s. 24.25 Fads3 n.s. n.s. 3.25 Herpud1 n.s. n.s. 4.59 Hspa9 n.s. n.s. 5.28 Leprotl1 n.s. n.s. 3.48 Steap1 n.s. n.s. 12.99 Genes downstream of ATF6 and/or PERK EDM5 EDM5/X MCDS Manf 2.13 n.s. 3.03 Pdia3 2.28 n.s. 2 Pdia4 2.52 n.s. 16 Genes downstream of Xbp1 and/or ATF6 EDM5 EDM5/X MCDS Creld2 2.84 -2.32 7.46 Dnajc3 3.98 -1.52 3.73 Grp94 1.86 n.s. 9.85 Hyou1 3.74 -1.66 4.59 Xbp1 2.07 -1.64 2.82 Genes downstream of synergy of ATF6 and Xbp1 EDM5 EDM5/X MCDS Derl1 n.s. n.s. 4.59 Derl2 n.s. n.s. 1.51 Edem1 n.s. n.s. 2.83 Edem2 n.s. n.s. n.s. Sel1l n.s. n.s. 4.59 Syvn1 n.s. n.s. 3.03 Vcp n.s. n.s. 1.51 https://doi.org/10.1371/journal.pgen.1008215.t001

PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008215 July 1, 2019 15 / 30 UPR is an essential component of genetic skeletal disease

mutant (Col10a1N617K vs. Col10a1WT) comparison. These were downregulated in mice expressing mutant matrilin-3 in both the presence and absence of XBP1, but were further downregulated upon ablation of Xbp1 indicating that they are ER-stress responsive and XBP1-dependent. Moreover, the transcriptomic comparison of EDM5 and MCDS growth plate chondrocytes suggested specific involvement of ATF6 and XBP1 signalling pathways in the UPR responses in the EDM5 mouse model, and an ATF6 and PERK-specific response in the MCDS cartilage (Table 1). In order to further confirm the lack of PERK involvement in EDM5 pathobiology, the UPR data obtained from the microarray analysis were verified by quantitative real-time RT-PCR analysis of Ire1, Atf6, Perk (Fig 7A). Interestingly, the expression levels of Ire1 and Perk were unchanged in all genotypes analysed, when compared to the wild type controls. Western blotting of total cartilage homogenates was used to verify the protein levels of IRE1, ATF6 and PERK (Fig 7B and 7C). Individual blots and loading controls are shown in S4 Fig. The results of densitometry analysis of the Western blots corresponded to the qPCR data and showed no difference in PERK and IRE1 protein levels across the genotypes, and an increase in ATF6 protein levels in EDM5 cartilage (1.7 fold), with a corresponding 1.4 fold increase in the amount of the active (cleaved) ATF6, further confirming the activation of the ATF6 signalling branch of the UPR in EDM5 tissue. ATF6 protein levels were also elevated in EDM5 cartilage lacking XBP1 (2.5 fold) when compared to wild type controls but were not statistically different when compared to EDM5 samples. We also assessed the expression levels of selected downstream targets of the IRE1, ATF6 and PERK signalling branches of the UPR in mRNA isolated from primary chondrocytes extracted from cartilage (Fig 8A). These genes included Pdia6 (downstream of Xbp1), Creld2 (regulated by Atf6 and Xbp1), Derl1 (downstream of synergistic actions of Xbp1 and Atf6), Grp94 (downstream of Atf6), Manf (regulated by Atf6 and Perk), and Ddit3 (downstream of Perk). Interestingly, whilst the levels of Xbp1 and Atf6 effectors were increased (Creld2 2.2-fold, Grp94 5.1 fold, Manf 1.6 fold, Pdia6 1.7 fold), the levels of Perk and of Ddit3, a pro- apoptotic gene specifically regulated by PERK, were not changed in the EDM5 tissues, further confirming the specific ATF6 and XBP1 involvement in EDM5 pathobiology. In addition, genes that are regulated or co-regulated by Xbp1 (Creld2, Derl1, Pdia6) were downregulated in EDM5 chondrocytes upon XBP1 removal, whereas genes downstream of Atf6 (Grp94, Manf) and Perk (Ddit3) were unchanged, further confirming our hypothesis. Moreover, Western blotting of total cartilage homogenates was used to verify the protein levels of PDIA6 (downstream of IRE1/XBP1), and CHOP (DDIT3) and ATF4 (both downstream of PERK), in Xbp1WT, Xbp1WT Matn3V194D, Xbp1Col2Δex2 and Xbp1Col2Δex2 Matn3V194D cartilage (Fig 8B). The results of densitometry analysis of the Western blots were consistent with the expression levels seen in the mRNA analysis and showed a 2.3 fold increase in PDIA6 levels in EDM5 cartilage and a decrease to wild type levels upon XBP1 deletion. Moreover, the densitometry analysis further confirmed the lack of involvement of the PERK branch of the UPR in the EDM5 disease signature and showed suppression of PERK mediated signals (both ATF4 and CHOP protein levels were decreased in EDM5 cartilage, 0.8 and 0.4 fold, respectively). Changes in the expression levels of ATF6/XBP1 regulated chaperone CRELD2, aggregation-related chaperone protein PDIA6 and proliferation regulator p58IPK (DNAJC3) were further verified in wild type and mutant cartilage growth plates by immunohistochemistry (Fig 8C). All three of these effectors were upregulated in the p.V194D Matn3 mutant cartilage and decreased upon deletion of XBP1, indicating XBP1-dependence of EDM5 UPR signalling. Interestingly, p58IPK has been shown to dampen PERK signalling and attenuate eIF2ɑ phosphorylation [40, 41]. Moreover, it has been shown that PERK can modulate Xbp1 expression and splicing in response to mutant protein aggregation [42, 43]. This effect of PERK signal on IRE1ɑ activity and splicing of Xbp1

PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008215 July 1, 2019 16 / 30 UPR is an essential component of genetic skeletal disease

PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008215 July 1, 2019 17 / 30 UPR is an essential component of genetic skeletal disease

Fig 7. (A) Real time RT-qPCR data verifying expression levels of Atf6, Ire1 and Perk using mRNA extracted from 5 day old chondrocytes (n = 3, Student t-test). No detectable changes in Perk expression confirm lack of involvement of the PERK signalling pathway in EDM5 pathobiology. (B) Densitometry measurement of 3 biological replicates of Western blotting for ATF6, IRE1 and PERK, showing no difference in IRE1 and PERK protein levels in EDM5 cartilage with or without XBP1. Interestingly, the amount of total and active (cleaved) ATF6 was dramatically increased in EDM5 cartilage compared to wild type controls and upon further XBP1 deletion. (C) Representative Western blotting images. Key: � P<0.05, �� P<0.01, �� P<0.001, scale bar 100μm. https://doi.org/10.1371/journal.pgen.1008215.g007

can be evidenced by a decreased ratio of alternatively spliced Xbp1 (Xbp1s) to unspliced Xbp1 (Xbp1u) in the EDM5 compared to the MCDS chondrocytes (S3 Fig).

Discussion The IRE1ɑ/XBP1 signalling branch is the most conserved branch of the UPR, essential in cellular response to the accumulation of misfolded proteins by regulating chaperone protein expression and ER-associated protein degradation (ERAD). Most likely due to the high protein secretory burden of chondrocytes and osteoblasts and the need for robust ER machinery during long bone growth [44, 45], this pathway is also important for skeletal development and was shown to regulate osteoblast differentiation in vitro [44] and chondrocyte proliferation and bone mineralisation in vivo [21]. XBP1 is the main effector of the pathway and Xbp1 gene is non-conventionally spliced by autophosphorylated IRE1ɑ following its dissociation from the main UPR sensor BiP. The spliced form (XBP1s) then translocates to the nucleus to act as a transcription factor and activates genes encoding chaperones and ERAD components [7]. The unspliced form (XBP1u) has a shorter half-life and can shuttle between the nucleus and the cytoplasm where it can aid in proteasomal degradation of XBP1s protein and/or the active form of ATF6 thereby acting as a regulator of UPR signalling [46, 47]. Xbp1s mRNA has been identified in the ‘disease signature’ of many conditions resulting from the ER retention of misfolded mutant protein [34] and in particular several diseases characterised by the formation of insoluble intracellular aggregates such as type II diabetes [20], Alzheimer’s disease [17, 35], Huntington’s disease [18, 19], metaphyseal chondrodysplasia type Schmidt (MCDS) [6, 36] and matrilin-3 related multiple epiphyseal dysplasia (EDM5) [5, 33], suggesting XBP1-dependent modulation of ER stress in these conditions. More specifically, it has previously been shown that MCDS and EDM5 share a common disease signature consistent with a classical UPR and defined by an upregulation of ATF6, alternative splicing of Xbp1 and an increase in Canx, Creld2, Derl3, Dnajc3, Hyou1, Manf, Pdia3, Pdia4, Pdia6 and Xbp1 gene expression [32]. It was therefore surprising that deletion of XBP1 in chondrocytes of a MCDS mouse model had no effect on the severity of the skeletal phenotype [31]. In contrast, deleting XBP1 from chondrocytes in the mouse model of EDM5 resulted in a dramatic increase in disease severity with significantly shorter limbs, deformed ribcages, severely dis- rupted cartilage growth plates and increased retention of mutant matrilin-3, suggesting an important role for XBP1 in response to abnormal protein aggregation in proliferating chondrocytes. Chaperone proteins involved in the processing of insoluble intracellular aggregates and already decreased in EDM5 chondrocytes (such as CRYAB, DNAJA1, DNAJA4, HSP1L and HSPA8), were further decreased following the deletion of XBP1. Interestingly, several of these were differentially expressed between the MCDS and EDM5 mice, indicating that the differentiation state of certain cells can influence their response to aggregation of mutant misfolded protein [36]. Moreover, several genes pertaining to proteasomal degradation (Derl1, Derl2, Edem1, Edem3) and autophagy (Atg2b, Atg4b, Atg5, Atg10, Atg12, Atg13) were upregulated in MCDS chondrocytes, but not EDM5 chondrocytes, indicating that despite toxic protein aggregation MCDS chondrocytes were able to upregulate crucial components of the degradation machinery. Interestingly, a recent study showed that the proteasomal and

PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008215 July 1, 2019 18 / 30 UPR is an essential component of genetic skeletal disease

PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008215 July 1, 2019 19 / 30 UPR is an essential component of genetic skeletal disease

Fig 8. (A) Real time RT-qPCR of genes downstream of XBP1, XBP1/ATF6 and PERK signalling. Creld2, Derl1 and Pdia6 (downstream of XBP1) follow an XBP1 dependent pattern of expression, and are upregulated in EDM5 cartilage and decreased upon deletion of XBP1. Grp94 and Manf (downstream of ATF6 and/or XBP1) appear to be ATF6-dependent in EDM5 cartilage and are not affected by lack of XBP1 in Xbp1Col2CreΔex2 Matn3V194D tissue. The mRNA levels of Ddit3 (Chop), a downstream effector of PERK, are not changed in any of the mouse lines analysed. (B) Densitometry measurement of 3 independent biological replicates of Western blotting for PDIA6 (downstream of XBP1), and ATF4 and DDIT3 (CHOP) (both downstream of PERK), showing a decrease in PERK effector levels in EDM5 cartilage. (C) Immunohistochemical staining of 3 week old cartilage growth plates showing a decrease in p58IPK, PDIA6 and CRELD2 staining (red) upon XBP1 deletion in the EDM5 cartilage. DAPI (blue) was used as a counterstain (blue). Key: � P<0.05, �� P<0.01, scale bar 100μm. https://doi.org/10.1371/journal.pgen.1008215.g008

autophagy inducer carbamazepine can enhance these responses and reduce the intracellular retention of mutant collagen X and restore long bone growth in MCDS mice [48]. The differences in the response to prolonged/chronic ER stress between proliferative and hypertrophic chondrocytes might stem from differentiation-specific variances in the basal levels of the three canonical ER stress pathways in the different zones of the cartilage growth plate. In physiological conditions the levels of ATF6ɑ and PERK are increased in hypertrophic chondrocytes compared to proliferative chondrocytes, whilst IRE1ɑ signalling appears more important in the proliferative zone of the growth plate (Fig 9A, [36, 49, 50]). It has been shown that the deletion of the ɑ or the ß isoform of ATF6 differentially affects the cartilage growth plate both in physiological conditions and following disease-associated ER stress, further confirming a modulation of the UPR through the differentiation state of a chondrocyte [51]. Inter- estingly, the levels of XBP1 are also increased in chondrocytes of the normal hypertrophic zone, potentially due to upregulated ATF6 signalling associated with hypertrophy [52, 53]. It is therefore not surprising that upon the induction of ER stress, ATF6ɑ and PERK were further upregulated in the hypertrophic MCDS chondrocytes. Xbp1 expression was induced in the hypertrophic cells of MCDS mice and Xbp1 was alternatively spliced, although the levels of IRE1ɑ did not increase (Fig 9A and 9B, [36]). Expression of genes downstream of ATF6 (including Xbp1 [54]) and PERK (Ddit3, Cebpb), as well as the expression of ERAD genes requiring the synergy of XBP1s and ATF6 signalling were upregulated in Col10a1N617K mice (Table 1, [55]). However, this was not sufficient in itself to trigger the ERAD pathway as evidenced by the increased intracellular retention of type X collagen in MCDS chondrocytes and the observation that MCDS mice lacking Xbp1 showed no increase in disease severity [31]. In contrast, ER/cell stress induced by the aggregation of misfolded mutant matrilin-3 protein in proliferative chondrocytes of EDM5 mice resulted in upregulation of Atf6a, and upregulation of protein levels of ATF6 and its downstream effectors (GRP94, MANF). Expression level of Ern1 (gene encoding IRE1ɑ) was not changed in EDM5 cartilage; however, the IRE1ɑ activity was enhanced, as evidenced by increased Xbp1 splicing and upregulation of XBP1 effectors (CRELD2, PDIA6). Interestingly, the levels of PERK and its downstream targets (ATF4, DDIT3) were not increased in EDM5 cartilage. ATF6 and XBP1 can form heterodi- mers and several UPR effectors can be modulated by either XBP1 or ATF6 or only by the actions of both [37, 55]. Interestingly, the expression of Dnajc3, which is downstream of XBP1 and ATF6, was increased in EDM5 chondrocytes and potentially further dampened PERK signalling [40, 41], as evidenced by a decrease in protein levels of PERK downstream effectors, ATF4 and DDIT3 in EDM5 cartilage. We therefore hypothesise that the UPR response in chondrocytes of the proliferative zone is primarily due to interplay between ATF6 and IRE1ɑ signalling and is not influenced by PERK. Therefore, the interplay between the XBP1 and ATF6 signalling is crucial for the pathobiology of aggregation-related diseases and that the modulation of the XBP1 pathway may present a promising therapeutic target. The potential role of PERK modulation of Xbp1 expression and Xbp1 splicing in response to mutant protein aggregation in hypertrophic cells is an interesting aspect that requires further investigation [42, 43]. The effect of PERK signalling on IRE1ɑ activity and splicing of

PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008215 July 1, 2019 20 / 30 UPR is an essential component of genetic skeletal disease

Fig 9. (A) Schematic explaining the basal levels of expression of the UPR sensors in the different zones of the cartilage growth plate and their differential expression upon the protein aggregation insult. (B) A schematic presenting the cross-talk of the UPR signalling branches upon misfolded protein aggregation

PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008215 July 1, 2019 21 / 30 UPR is an essential component of genetic skeletal disease

in the proliferative (EDM5) and hypertrophic (MCDS) zones. (C) STRING visualisation of XBP1 modulated genes differentially expressed in the 5 day old EDM5 cartilage compared to the wild type control showing the postulated “aggregosome”. https://doi.org/10.1371/journal.pgen.1008215.g009

Xbp1 can be evidenced by a decreased ratio of Xbp1s:Xbp1u in the EDM5 compared to the MCDS chondrocytes. It is therefore interesting to speculate that inducing the upregulation of PERK signalling in EDM5 chondrocytes may lead to a preferential upregulation of XBP1-dependent ERAD or autophagy [43]. The upregulation of many ERAD and autophagy components in the MCDS cartilage, and the ability of carbamazepine to induce degradation of mutant collagen X, may also be a result of the hypertrophic chondrocytes being “ERAD primed” through their differentiation process (Fig 9B, [36, 56, 57]). Several of the XBP1-dependent genes identified in our EDM5 study play a role in an intracellular pathway for the removal of xenobiotics and lipophilic substances. In particular Abcb1, Ugta1a1 and Hao1 were upregulated in the EDM5 chondrocytes and downregulated upon removal of Xbp1, indicating an alternative XBP1-dependent pathway triggered by the aggregation of mutant protein. STRING visualisation of the UPR genes differentially expressed in EDM5 cartilage and modulated by XBP1 reveals a potential experimentally supported “aggregosome” (Fig 9C). Interestingly, this disease signature is similar to Alzheimer’s disease (AD) models with intracellular retention of amyloid-like deposits. Specifically, Creld2, Dnajc3, Manf, Pdia3 and Pdia6 are upregulated both in EDM5 and in AD, and Cryab, Dnaja4 and Hsph1 are downregulated in both conditions whilst ERAD components are not affected [58]. CRYAB and HSPH1 are both chaperone proteins that prevent aggregation of mutant proteins and defects in CRYAB have been associated with Huntington’s disease and Alzheimer’s disease, where its levels decrease in age-dependent manner [59–61]. It is therefore interesting to speculate that these two chaperones could represent potential therapeutic targets for EDM5 and other protein aggregation related diseases. In fact, overexpression of CRYAB in a mouse model of Huntington’s disease was shown to be neuroprotective and reduced the size of aggre- gate inclusions in the affected brains [62]. Finally, a polymorphism in the Xbp1 promoter is one of the risk factors for Alzheimer’s disease [17] and the pathobiology of protein aggregation conditions such as Alzheimer’s and Huntington’s disease can be regulated via a modulation of the IRE1ɑ-XBP1 branch of the UPR [18, 19, 35]. Our data suggests that the pathobiology of EDM5 is governed by the tight regulation between the IRE1ɑ and ATF6 signalling and a balance between the spliced and unspliced forms of Xbp1. It is therefore interesting to speculate that modifying the XBP1 signalling pathway via chemical or genetic intervention might present a therapeutic avenue for a broader range of protein aggregation diseases, leading to an increase in ERAD or autophagy, or an increased sensitivity to degradation-inducing therapies such as carbamazepine treatment. Sev- eral studies in normal cells, cancer and disease models have previously identified chemicals that activate or block the RNAse and/or kinase activity of IRE1ɑ and modulate the Xbp1s: Xbp1u ratio as potential therapeutic modifiers of the IRE1ɑ/XBP1 signalling pathway [63–66]. This study further confirms the importance of this pathway in degradation of insoluble intracellular aggregates and offers a novel therapeutic avenue that could be applicable to a broader range of aggregation conditions.

Materials and methods Generation of mouse lines Xbp1WT Matn3V194D mice [33] were crossed with Xbp1Col2CreΔex2 mice [21] in which Xbp1 mRNA is inactivated by the Col2a1 promoter-driven Cre recombinase-mediated deletion of

PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008215 July 1, 2019 22 / 30 UPR is an essential component of genetic skeletal disease

exon 2 to generate the compound mutant mouse line, Xbp1Col2CreΔex2 Matn3V194D. All the mice were generated on the C57BL6/J background to control for the background effects. Gen- otyping was performed as previously described and RT-PCR and sequencing were performed on cartilage RNA to confirm deletion of Xbp1 exon 2 in mutant chondrocytes.

Ethics statement All experiments were approved by the University of Manchester Animal Ethical Review Group and performed in compliance with the Scientific Procedures Act of 1986 and the relevant Home Office (under PPL 40/2884 and PPL60/04525) and Institutional regulations governing animal breeding and handling.

Bone measurements Mice of different genotypes (>5 per age per genotype) were sacrificed at 3, 6 and 9 weeks of age and X-rayed using Faxitron MX-20 X-ray machine. The bones were measured using Fiji ImageJ platform (National Institutes of Health, Bethesda, Maryland, USA; [67]) and one way ANOVA and Student t-test were applied for statistical analysis.

Histology and immunohistochemistry Mice were sacrificed at 3 weeks of age, hindlimbs were dissected and fixed in either PFA (histology) or 95% ethanol 5% acetic acid (immunohistochemistry) for 48h in 4˚C. The limbs were then decalcified in 20% EDTA pH 7.4 over 2 weeks, wax embedded and cut into 6μm sections. Haematoxylin/eosin (H&E) staining was used to visualise the general morphology of the tissue, using the automated Thermo Shandon stainer. Immunohistochemistry and BrdU labelling (measurement of cell proliferation) were performed as described previously [30]. Primary antibodies were used at a dilution of 1:500 (ER stress: BiP, GRP94 and CRELD2 from R&D Systems; PDIA6 from Abcam, p58IPK from Santa Cruz Biotechnology; ECM: type I collagen, type II collagen from Abcam; matrilin-3 from R&D Systems; type X collagen [68]). BrdU labelled cells were counted using the Watershed algorithm on the Fiji ImageJ platform (National Institutes of Health, Bethesda, Maryland, USA; [67]) and presented as percentage of total cells in the proliferative zone and On-Way ANOVA was used for statistical analysis of data.

TUNEL assay TUNEL assay was performed on PFA fixed sections of 3 week old limbs using the Promega Dead-End Fluorimetric Kit as previously [33]. The samples were unmasked using citric buffer boil instead of proteinase K unmasking, which can generate false positives [69]. Positive cells labelled with FITC were counted using the Watershed algorithm on Fiji ImageJ platform (National Institutes of Health, Bethesda, Maryland, USA; [67]) and presented as percentage of all (DAPI stained) cells in selected zones of the growth plate. One-Way ANOVA was used for statistical analysis of the data.

Isolation of costal chondrocytes Costal chondrocytes were isolated from pooled litters of 5 day-old mice as described previously [5]. For RNA analysis, the cell pellet was resuspended in TRIzol reagent (Invitrogen), flash frozen and stored at -80˚C until RNA extraction. For protein analysis, chondrocyte aliquots of 1.5 x 105 cells were re-suspended in SDS loading buffer and frozen at -20˚C until analysis.

PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008215 July 1, 2019 23 / 30 UPR is an essential component of genetic skeletal disease

Microarray analysis of rib chondrocytes RNA was extracted from isolated chondrocytes using Trizol reagent, according to the manu- facturer’s protocol (Life Technologies). Wild type and mutant RNA was pooled from 3 separate extractions, and the submitted to the Genomic Technologies Core Facility, University of Manchester for analysis. RNA integrity was analysed on the 2100 Bioanalyser (Agilent Tech- nologies). A GeneChip 3’ expression assay (Mouse430_2 Affymetrix) was used to analyse gene expression. Quality control checks for control hybridizations were performed using Microar- ray Suite 5. PPLR is an R package that detects differential gene expression by including probe- level measurement error and calculating the probability of positive log-ratio (PPLR). The differentially expressed (over 1.5 fold change, PPLR of 1.0 and 0.0) genes were subjected to functional annotation analysis using Database for Annotation, Visualisation and Integrated Discovery (DAVID) software [70, 71]. This analysis assigned significantly up/downregulated genes into structural, compartmental and functional-related clusters (GOTERMS). Pathway analysis of the significantly changed genes in the Xbp1Col2CreΔex2 Matn3V194D vs Xbp1WT Matn3V194D microarray data was performed using the iPathwayGuide platform (Advaita Corp; http://www.advaitabio.com/ipathwayguide). This software analysis tool implements the Impact Analysis approach that takes into consideration the direction and type of all signals on a pathway [72].

Data availability The full datasets are available from the NCBI Gene Expression Omnibus (GEO), accession number GSE120308 (http://www.ncbi.nlm.nih.gov/geo/).

Quantitative RT-PCR analysis of rib chondrocytes Microarray data were verified by quantitative real time PCR. Briefly, First-strand cDNA was synthesised using random hexamer primers and the GoScript Reverse Transcriptase System (Promega), and qPCR was performed using the SYBR green PCR protocol (Applied Biosys- tems) on the Chromo4 real-time PCR system (Bio-Rad). Primer sequences are presented in S5 Table. Each experiment included ‘no template’ controls, was run in duplicate and had an 18S RNA control. Each independent experiment was repeated three times, and the results were analysed by independent-samples t-test.

Western blotting of cartilage homogenates Tissue homogenates were prepared by homogenising liquid nitrogen frozen 3 week old femoral head tissue in PBS using a microdismembranator for 2 min at 2,000rpm (Sartorius Ltd). Samples (30μg of total protein) were denatured at 95˚C for 5 min in SDS-PAGE loading buffer containing 100 mM dithiotreitol (DTT. Proteins were separated by SDS-PAGE using Novex NuPAGE 4–12% Bis Tris precast gels in MES running buffer (Fisher Scientific) at 200 V for 60 minutes and electroblotted onto a nitrocellulose membrane for 1 hour at 30 V. Gel loading was assessed using REVERT Total Protein Stain (LI-COR Biosciences) according to manufac- turer’s instructions. Membranes were blocked in 3% Milk in PBS-T, incubated with primary antibodies (1:100, ATF4 118155 (Cell Signalling Technology Inc.), ATF6 70B1413.1 (Enzo Life Sciences Inc.), DDIT3 ab11419, IRE1 ab37073, PDIA6 ab154820 (Abcam), PERK C33E10 (Cell Signalling Technology Inc.)) for 1 hour at room temperature, and probed with the appro- priate LI-COR IRDye secondary antibody at a concentration of 1:5,000 for 1 hour. Blots were imaged on the LI-COR Odyssey CLx Imaging System. Densitometry quantification was performed using LI-COR proprietary software and verified by ImageJ. The densitometry data was

PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008215 July 1, 2019 24 / 30 UPR is an essential component of genetic skeletal disease

normalised to the total protein stain to account for protein loading. The analysis was undertaken independently by two different researchers, the data was then normalised to wild type protein levels and statistically analysed

Supporting information S1 Fig. X rays of one year old Xbp1WT Matn3V194D and Xbp1Col2CreΔex2 Matn3V194D mice showing a striking limb deformity and rotation, potentially due to the pronounced hip dysplasia as well as the constricted bell shaped ribcage characteristic of the Xbp1Col2CreΔex2 Matn3V194D line. (TIF) S2 Fig. Fluorescence imaging of Congo Red staining of mouse growth plates at 3 weeks of age showing potential amyloid-like deposits in the Xbp1WT Matn3V194D and Xbp1Col2Cre- Δex2 Matn3V194D cartilage but not rdw cartilage (positive control). Scale bar 100μm. (TIF) S3 Fig. (A) A heat map generated for top 200 upregulated (in red) and downregulated (in blue) probes in the Matn3V194D vs Matn3WT (EDM5) analysis showing a high percentage of XBP1-dependent genes (upregulated in the EDM5, downregulated in Xbp1 null cartilage and in EDM5 lacking Xbp1). (B) Venn diagram analysis showing a comparison of all differentially expressed genes in the compared mouse models. (C) Venn diagram showing differential gene expression in the hypertrophic zone of Xbp1Col2CreΔex2 vs Xbp1WT cartilage compared to the expression profile of the entire growth plate. (D) Venn diagram comparison of the differential gene expression between the MCD and EDM5 mouse models. (E) RT-PCR quantification of the relative levels of Xbp1u:Xbp1s in 5 day old chondrocytes showing higher availability of the Xbp1s in the MCDS mouse model compared to the EDM5 mouse (n = 3, Student t-test). (TIF) S4 Fig. (A) Total protein stain showing protein loading and individual Western blotting for ATF6, IRE1 and PERK on 3 independent biological replicates of whole femoral head cartilage homogenates at 3 weeks of age. B) Total protein stain showing protein loading and individual Western blotting for ATF4, DDIT3, and PDIA6 on 3 independent biological replicates of whole femoral cartilage homogenates at 3 weeks of age. Key: uATF6 –uncleaved ATF6, cATF6 –cleaved (active) ATF6 protein. (TIF) S1 Table. Significantly upregulated cell proliferation, migration and cellular response to ER stress genes in the Xbp1WT Matn3V194D vs Xbp1WT microarray comparison. (DOCX) S2 Table. List of 57 genes (75 probes) significantly downregulated in the Xbp1WT Matn3V194D vs Xbp1WT comparison and further downregulated in the Xbp1Col2CreΔex2 Matn3V194D vs Xbp1WT Matn3V194D comparison. (DOCX) S3 Table. 150 genes significantly upregulated in the Xbp1WT Matn3V194D vs Xbp1WT analysis and downregulated upon removal of Xbp1 (Xbp1Col2CreΔex2 Matn3V194D vs Xbp1WT Matn3V194D). (DOCX)

PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008215 July 1, 2019 25 / 30 UPR is an essential component of genetic skeletal disease

S4 Table. 79 genes (97 probes) downregulated in Xbp1WT Matn3V194D vs Xbp1WT analysis and upregulated in Xbp1Col2CreΔex2 Matn3V194D vs Xbp1WT Matn3V194D dataset. (DOCX) S5 Table. Quantitative RT-PCR primer sequences. (DOCX)

Acknowledgments We thank Prof Laurie H. Glimcher (Dana-Farber Cancer Institute, Boston, USA) for sharing the Xbp1 conditional knock-out line, without which this study would not be possible. We would also like to acknowledge the Manchester University Transgenic, Transcriptomic and Proteomic Core Facilities for their help with animal maintenance, sample processing and data analysis and the Newcastle University Comparative Biology Centre for their help with animal husbandry during the project.

Author Contributions Conceptualization: Katarzyna A. Piro´g, Claire L. Hartley, Raymond P. Boot-Handford, Michael D. Briggs. Data curation: Katarzyna A. Piro´g, Ella P. Dennis, Robert M. Jackson, Jamie Soul. Formal analysis: Katarzyna A. Piro´g, Ella P. Dennis, Robert M. Jackson, Jamie Soul. Funding acquisition: Katarzyna A. Piro´g, Jean-Marc Schwartz, John F. Bateman, Raymond P. Boot-Handford, Michael D. Briggs. Investigation: Katarzyna A. Piro´g, Ella P. Dennis, Claire L. Hartley. Methodology: Katarzyna A. Piro´g, Jamie Soul. Project administration: Katarzyna A. Piro´g. Resources: John F. Bateman. Supervision: Jean-Marc Schwartz, John F. Bateman, Raymond P. Boot-Handford, Michael D. Briggs. Validation: Katarzyna A. Piro´g, Ella P. Dennis, Claire L. Hartley, Robert M. Jackson. Visualization: Katarzyna A. Piro´g. Writing – original draft: Katarzyna A. Piro´g. Writing – review & editing: Katarzyna A. Piro´g, John F. Bateman, Raymond P. Boot-Hand- ford, Michael D. Briggs.

References 1. Kaufman RJ. Orchestrating the unfolded protein response in health and disease. J Clin Invest. 2002; 110(10):1389±98. https://doi.org/10.1172/JCI16886 PMID: 12438434 2. Song B, Scheuner D, Ron D, Pennathur S, Kaufman R. Chop deletion reduces oxidative stress, improves β cell function, and promotes cell survival in multiple mouse models of diabetes. J Clin Invest. 2008; 118(10):3378±89. https://doi.org/10.1172/JCI34587 PMID: 18776938 3. McLaughlin M, Karim S, Montague P, Barrie J, Kirkham D, Griffiths I, et al. Genetic background influences UPR but not PLP processing in the rumpshaker model of PMD/SPG2. Neurochem Res. 2007; 32 (2):167±76. https://doi.org/10.1007/s11064-006-9122-y PMID: 16944321 4. Porcellotti S, Fanelli F, Fracassi A, Sepe S, Cecconi F, Bernardi C, et al. Oxidative Stress during the Progression of beta-Amyloid Pathology in the Neocortex of the Tg2576 Mouse Model of Alzheimer's

PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008215 July 1, 2019 26 / 30 UPR is an essential component of genetic skeletal disease

Disease. Oxidative medicine and cellular longevity. 2015; 2015:967203. https://doi.org/10.1155/2015/ 967203 PMID: 25973140 5. Nundlall S, Rajpar MH, Bell PA, Clowes C, Zeeff LA, Gardner B, et al. An unfolded protein response is the initial cellular response to the expression of mutant matrilin-3 in a mouse model of multiple epiphyseal dysplasia. Cell Stress Chaperones. 2010; 15(6):835±49. https://doi.org/10.1007/s12192-010- 0193-y PMID: 20428984 6. Rajpar H, McDermott B, Kung L, Eardley R, Knowles L, Heeran M, et al. Targeted induction of endoplasmic reticulum stress induces cartilage pathology. PLoS Genet. 2009; 5(10):e1000691. https://doi. org/10.1371/journal.pgen.1000691 PMID: 19834559 7. Boot-Handford RP, Briggs MD. The unfolded protein response and its relevance to connective tissue diseases. Cell and tissue research. 2010; 339(1):197±211. https://doi.org/10.1007/s00441-009-0877-8 PMID: 19851784 8. Rutkowski DT, Kaufman RJ. A trip to the ER: coping with stress. Trends in Cell Biology. 2004; 14:20±8. PMID: 14729177 9. Zhang P, McGrath B, Li S, Frank A, Zambito F, Reinert J, et al. The PERK Eukaryotic Initiation Factor 2α Kinase Is Required for the Development of the Skeletal System, Postnatal Growth, and the Function and Viability of the Pancreas. Mol Cell Biol. 2002; 22(11):3864±74. https://doi.org/10.1128/MCB.22.11. 3864-3874.2002 PMID: 11997520 10. Averous J, Bruhat A, Jousse C, Carraro V, Thiel G, Fafournoux P. Induction of CHOP expression by amino acid limitation requires both ATF4 expression and ATF2 phosphorylation. J Biol Chem. 2004; 279(7):5288±97. https://doi.org/10.1074/jbc.M311862200 PMID: 14630918 11. Marciniak S, Yun C, Oyadomari S, Novoa I, Zhang Y, Jungreis R, et al. CHOP induces death by promoting protein synthesis and oxidation in the stressed endoplasmic reticulum. Genes Dev. 2004; 18(24). 12. Guo F, Xiong Z, Lu X, Ye M, Han X, Jiang R. ATF6 upregulates XBP1S and inhibits ER stress-mediated apoptosis in osteoarthritis cartilage. Cell Signal. 2014; 26(2):332±42. https://doi.org/10.1016/j.cellsig. 2013.11.018 PMID: 24269637 13. Zhang K, Wong H, Song B, Miller CN, Scheuner D, Kaufman RJ. The unfolded protein response sensor IRE1alpha is required at 2 distinct steps in B cell lymphopoiesis. J Clin Invest. 2005; 115:268±81. https://doi.org/10.1172/JCI21848 PMID: 15690081 14. Han X, Zhou J, Zhang P, Jiang R, Li M, Xia F, et al. IRE1α dissociates with BiP and inhibits ER stress- mediated apoptosis in cartilage development. Cell Signal. 2013; 25(11):2136±46. https://doi.org/10. 1016/j.cellsig.2013.06.011 PMID: 23816533 15. Gardner B, Walter P. Unfolded proteins are Ire1-activating ligands that directly induce the unfolded protein response. Science. 2011; 333(6051):1891±4. https://doi.org/10.1126/science.1209126 PMID: 21852455 16. Yoo Y, Byun K, Kang T, Bayarsaikhan D, Kim JY, Oh S, et al. Amyloid-beta-activated human microglial cells through ER-resident proteins. Journal of Proteome Research. 2015; 14(1):214±23. https://doi.org/ 10.1021/pr500926r PMID: 25384129 17. Duran-Aniotz C, Cornejo VH, Espinoza S, Ardiles AÂ O, Medinas DB, Salazar C, et al. IRE1 signaling exacerbates Alzheimer's disease pathogenesis. Acta Neuropathologica. 2017; 134(3):489±506. https:// doi.org/10.1007/s00401-017-1694-x PMID: 28341998 18. Vidal RL, Figueroa A, Court FA, Thielen P, Molina C, Wirth C, et al. Targeting the UPR transcription factor XBP1 protects against Huntington's disease through the regulation of FoxO1 and autophagy. Human molecular genetics. 2012; 21(10):2245±62. https://doi.org/10.1093/hmg/dds040 PMID: 22337954 19. Zuleta A, Vidal RL, Armentano D, Parsons G, Hetz C. AAV-mediated delivery of the transcription factor XBP1s into the striatum reduces mutant Huntingtin aggregation in a mouse model of Huntington's disease. Biochemical and Biophysical Research Communications. 2012; 420(3):558±63. https://doi.org/ 10.1016/j.bbrc.2012.03.033 PMID: 22445760 20. Akiyama M, Liew CW, Lu S, Hu J, Martinez R, Hambro B, et al. X-Box Binding Protein 1 Is Essential for Insulin Regulation of Pancreatic α-Cell Function. Diabetes. 2013; 62(7):2439±49. https://doi.org/10. 2337/db12-1747 PMID: 23493568 21. Cameron TL, Gresshoff IL, Bell KM, PiroÂg KA, Sampurno L, Hartley CL, et al. Cartilage-specific ablation of XBP1 signaling in mouse results in a chondrodysplasia characterized by reduced chondrocyte proliferation and delayed cartilage maturation and mineralization. Osteoarthritis and Cartilage. 2015; 23 (4):661±70. https://doi.org/10.1016/j.joca.2015.01.001 PMID: 25600960 22. Briggs MD, Wright MJ, Mortier GR. Multiple Epiphyseal Dysplasia, Dominant. In: Pagon RA, Adam MP, Bird TD, Dolan CR, Fong CT, Smith RJH, et al., editors. GeneReviews(R). Seattle WA: University of Washington, Seattle; 1993.

PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008215 July 1, 2019 27 / 30 UPR is an essential component of genetic skeletal disease

23. Chapman KL, Briggs MD. Review: Clinical variability and genetic heterogeneity in multiple epiphyseal dysplasia. Pediatric Pathology and Molecular Medicine. 2003; 22:53±75. PMID: 12687890 24. Nicolae C, Ko YP, Miosge N, Niehoff A, Studer D, Enggist L, et al. Abnormal collagen fibrils in cartilage of matrilin-1/matrilin-3-deficient mice.. J Biol Chem. 2007; 282:22163±75. https://doi.org/10.1074/jbc. M610994200 PMID: 17502381 25. Klatt AR, Nitsche DP, Kobbe B, Moergelin M, Paulsson M, Wagener R. Molecular structure and tissue distribution of matrilin-3, a filament forming extracellular matrix protein expressed during skeletal development. J Biol Chem. 2000; 275:3999±4006. https://doi.org/10.1074/jbc.275.6.3999 PMID: 10660556 26. Rambaran RN, Serpell LC. Amyloid fibrils: Abnormal protein assembly. Prion. 2008; 2(3):112±7. https:// doi.org/10.4161/pri.2.3.7488 PMID: 19158505 27. Otten C, Wagener R, Paulsson M, Zaucke F. Matrilin-3 mutations that cause chondrodysplasias inter- fere with protein trafficking while a mutation associated with hand osteoarthritis does not. J Med Genet. 2005; 42(10):774±9. https://doi.org/10.1136/jmg.2004.029462 PMID: 16199550 28. Jackson GC, Barker FS, Jakkula E, Czarny-Ratajczak M, Makitie O, Cole WG, et al. Missense mutations in the beta strands of the single A-domain of matrilin-3 result in multiple epiphyseal dysplasia. J Med Genet. 2004; 41(1):52±9. https://doi.org/10.1136/jmg.2003.011429 PMID: 14729835 29. Bell PA, Wagener R, Zaucke F, Koch M, Selley J, Warwood S, et al. Analysis of the cartilage proteome from three different mouse models of genetic skeletal diseases reveals common and discrete disease signatures. Biol Open. 2013; 2(8):802±11. https://doi.org/10.1242/bio.20135280 PMID: 23951406 30. Pirog-Garcia KA, Meadows RS, Knowles L, Heinegard D, Thornton DJ, Kadler KE, et al. Reduced cell proliferation and increased apoptosis are significant pathological mechanisms in a murine model of mild pseudoachondroplasia resulting from a mutation in the C-terminal domain of COMP. Human molecular genetics. 2007; 16(17):2072±88. https://doi.org/10.1093/hmg/ddm155 PMID: 17588960 31. Cameron TL, Bell KM, Gresshoff IL, Sampurno L, Mullan L, Ermann J, et al. XBP1-Independent UPR Pathways Suppress C/EBP-β Mediated Chondrocyte Differentiation in ER-Stress Related Skeletal Dis- ease. PLOS Genetics. 2015; 11(9):e1005505. https://doi.org/10.1371/journal.pgen.1005505 PMID: 26372225 32. Hartley CL, Edwards S, Mullan L, Bell PA, Fresquet M, Boot-Handford RP, et al. Armet/Manf and Creld2 are components of a specialized ER stress response provoked by inappropriate formation of disulphide bonds: implications for genetic skeletal diseases. Human molecular genetics. 2013; 22 (25):5262±75. https://doi.org/10.1093/hmg/ddt383 PMID: 23956175 33. Leighton MP, Nundlall S, Starborg T, Meadows RS, Suleman F, Knowles L, et al. Decreased chondrocyte proliferation and dysregulated apoptosis in the cartilage growth plate are key features of a murine model of epiphyseal dysplasia caused by a matn3 mutation. Human molecular genetics. 2007; 16 (14):1728±41. https://doi.org/10.1093/hmg/ddm121 PMID: 17517694 34. Acosta-Alvear D, Zhou Y, Blais A, Tsikitis M, Lents NH, Arias C, et al. XBP1 Controls Diverse Cell Type- and Condition-Specific Transcriptional Regulatory Networks. Molecular Cell. 2007; 27(1):53±66. https:// doi.org/10.1016/j.molcel.2007.06.011 PMID: 17612490 35. Meng T, Cao Q, Lei P, Bush AI, Xiang Q, Su Z, et al. Tat-haFGF14-154 Upregulates ADAM10 to Attenu- ate the Alzheimer Phenotype of APP/PS1 Mice through the PI3K-CREB-IRE1α/XBP1 Pathway. Mol Ther Nucleic Acids. 2017:439±52. https://doi.org/10.1016/j.omtn.2017.05.004 PMID: 28624220 36. Cameron TL, Bell KM, Tatarczuch L, Mackie EJ, Rajpar MH, McDermott BT, et al. Transcriptional Profil- ing of Chondrodysplasia Growth Plate Cartilage Reveals Adaptive ER-Stress Networks That Allow Sur- vival but Disrupt Hypertrophy. PLOS ONE. 2011; 6(9):e24600. https://doi.org/10.1371/journal.pone. 0024600 PMID: 21935428 37. Lee AH, Iwakoshi NN, Glimcher LH. Xbp1 Regulates a Subset of Endoplasmic Reticulum Resident Chaperone Genes in the Unfolded Protein Response. Mol Cell Biol. 2003; 23:7448±59. https://doi.org/ 10.1128/MCB.23.21.7448-7459.2003 PMID: 14559994 38. Fels D, Koumenis C. The PERK/eIF2alpha/ATF4 module of the UPR in hypoxia resistance and tumor growth. Cancer Biol Ther. 2006; 5(7):723±8. https://doi.org/10.4161/cbt.5.7.2967 PMID: 16861899 39. Schroeder M. Endoplasmic reticulum stress responses. Cell Mol Life Sci. 2008; 65:862±94. https://doi. org/10.1007/s00018-007-7383-5 PMID: 18038217 40. Yan W, Frank CL, Korth MJ, Sopher BL, Novoa I, Ron D, et al. Control of PERK eIF2alpha kinase activity by the endoplasmic reticulum stress-induced molecular chaperone P58IPK. Proc Natl Acad Sci U S A. 2002; 99(25):15920±5. https://doi.org/10.1073/pnas.252341799 PMID: 12446838 41. van Huizen R, Martindale JL, Gorospe M, Holbrook NJ. P58IPK, a novel endoplasmic reticulum stress- inducible protein and potential negative regulator of eIF2alpha signaling. J Biol Chem. 2003; 278 (18):15558±64. https://doi.org/10.1074/jbc.M212074200 PMID: 12601012

PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008215 July 1, 2019 28 / 30 UPR is an essential component of genetic skeletal disease

42. Fusakio ME, Willy JA, Wang Y, Mirek ET, Al Baghdadi RJ, Adams CM, et al. Transcription factor ATF4 directs basal and stress-induced gene expression in the unfolded protein response and cholesterol metabolism in the liver. Molecular biology of the cell. 2016; 27(9):1536±51. https://doi.org/10.1091/mbc. E16-01-0039 PMID: 26960794 43. Tsuru A, Imai Y, Saito M, Kohno K. Novel mechanism of enhancing IRE1alpha-XBP1 signalling via the PERK-ATF4 pathway. Scientific reports. 2016; 6:24217. https://doi.org/10.1038/srep24217 PMID: 27052593 44. Tohmonda T, Miyauchi Y, Ghosh R, Yoda M, Uchikawa S, Takito J, et al. The IRE1α-XBP1 pathway is essential for osteoblast differentiation through promoting transcription of Osterix. EMBO Rep. 2011; 12 (5):451±7. https://doi.org/10.1038/embor.2011.34 PMID: 21415858 45. Pacifici M, Iozzo RV. Remodeling of the rough endoplasmic reticulum during stimulation of procollagen secretion by ascorbic acid in cultured chondrocytes. A biochemical and morphological study. J Biol Chem. 1988; 263(5):2483±92. PMID: 3276690 46. Yoshida H, Oku M, Suzuki M, Mori K. pXBP1(U) encoded in XBP1 pre-mRNA negatively regulates unfolded protein response activator pXBP1(S) in mammalian ER stress response. J Cell Biol. 2006; 172(4):565±75. https://doi.org/10.1083/jcb.200508145 PMID: 16461360 47. Horimoto S, Ninagawa S, Okada T, Koba H, Sugimoto T, Kamiya Y, et al. The Unfolded Protein Response Transducer ATF6 Represents a Novel Transmembrane-type Endoplasmic Reticulum-associated Degradation Substrate Requiring Both Mannose Trimming and SEL1L Protein. The Journal of Biological Chemistry. 2013; 288(44):31517±27. https://doi.org/10.1074/jbc.M113.476010 PMID: 24043630 48. Mullan LA, Mularczyk EJ, Kung LH, Forouhan M, Wragg JM, Goodacre R, et al. Increased intracellular proteolysis reduces disease severity in an ER stress-associated dwarfism. J Clin Invest. 2017; 127 (10):3861±5. https://doi.org/10.1172/JCI93094 PMID: 28920921 49. Xiong Z, Jiang R, Zhang P, Han X, Guo FJ. Transmission of ER stress response by ATF6 promotes endochondral bone growth. Journal of orthopaedic surgery and research. 2015; 10:141. https://doi.org/ 10.1186/s13018-015-0284-7 PMID: 26374329 50. Kang X, Yang W, Feng D, Jin X, Ma Z, Qian Z, et al. Cartilage-Specific Autophagy Deficiency Promotes ER Stress and Impairs Chondrogenesis in PERK-ATF4-CHOP-Dependent Manner. J Bone Miner Res. 2017; 32(10):2128±41. Epub 2017/03/18. https://doi.org/10.1002/jbmr.3134 PMID: 28304100 51. Forouhan M, Mori K, Boot-Handford RP. Paradoxical roles of ATF6alpha and ATF6beta in modulating disease severity caused by mutations in collagen X. Matrix biology : journal of the International Society for Matrix Biology. 2018; 70:50±71. 52. Belluoccio D, Bernardo BC, Rowley L, Bateman JF. A microarray approach for comparative expression profiling of the discrete maturation zones of mouse growth plate cartilage. Biochim Biophys Acta. 2008; 1779(5):330±40. https://doi.org/10.1016/j.bbagrm.2008.02.010 PMID: 18374667 53. Liu Y, Zhou J, Zhao W, Li X, Jiang R, Liu C, et al. XBP1S associates with RUNX2 and regulates chondrocyte hypertrophy. J Biol Chem. 2012; 287(41):34500±13. https://doi.org/10.1074/jbc.M112.385922 PMID: 22865880 54. Yoshida H, Matsui T, Yamamoto A, Okada T, Mori K. XBP1 mRNA is induced by ATF6 and spliced by IRE1 in response to ER stress to produce a highly active transcription factor. Cell. 2001; 107(7):881±91. PMID: 11779464 55. Shoulders Matthew D, Ryno Lisa M, Genereux Joseph C, Moresco James J, Tu Patricia G, Wu C, et al. Stress-Independent Activation of XBP1s and/or ATF6 Reveals Three Functionally Diverse ER Proteos- tasis Environments. Cell Reports. 2013; 3(4):1279±92. https://doi.org/10.1016/j.celrep.2013.03.024 PMID: 23583182 56. Kishino A, Hayashi K, Hidai C, Masuda T, Nomura Y, Oshima T. XBP1-FoxO1 interaction regulates ER stress-induced autophagy in auditory cells. Scientific reports. 2017; 7(1):4442. https://doi.org/10.1038/ s41598-017-02960-1 PMID: 28667325 57. Zhao Y, Li X, Ma K, Yang J, Zhou J, Fu W, et al. The axis of MAPK1/3-XBP1u-FOXO1 controls autopha- gic dynamics in cancer cells. Autophagy. 2013; 9(5):794±6. https://doi.org/10.4161/auto.23918 PMID: 23426330 58. Ghribi O. The role of the endoplasmic reticulum in the accumulation of beta-amyloid peptide in Alzhei- mer's disease. Current molecular medicine. 2006; 6(1):119±33. PMID: 16472119 59. Ojha J, Karmegam RV, Masilamoni JG, Terry AV, Cashikar AG. Behavioral defects in chaperone-deficient Alzheimer's disease model mice. PLoS One. 2011; 6(2):e16550. https://doi.org/10.1371/journal. pone.0016550 PMID: 21379584 60. Yamagishi N, Yokota M, Yasuda K, Saito Y, Nagata K, Hatayama T. Characterization of stress sensitivity and chaperone activity of Hsp105 in mammalian cells. Biochem Biophys Res Commun. 2011; 409 (1):90±5. https://doi.org/10.1016/j.bbrc.2011.04.114 PMID: 21557931

PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008215 July 1, 2019 29 / 30 UPR is an essential component of genetic skeletal disease

61. Yamagishi N, Goto K, Nakagawa S, Saito Y, Hatayama T. Hsp105 reduces the protein aggregation and cytotoxicity by expanded-polyglutamine proteins through the induction of Hsp70. Experimental cell research. 2010; 316(15):2424±33. https://doi.org/10.1016/j.yexcr.2010.06.003 PMID: 20542028 62. Oliveira AO, Osmand A, Outeiro TF, Muchowski PJ, Finkbeiner S. αB-Crystallin overexpression in astrocytes modulates the phenotype of the BACHD mouse model of Huntington's disease. Human Molecular Genetics. 2016; 25(9):1677±89. https://doi.org/10.1093/hmg/ddw028 PMID: 26920069 63. Volkmann K, Lucas JL, Vuga D, Wang X, Brumm D, Stiles C, et al. Potent and selective inhibitors of the inositol-requiring enzyme 1 endoribonuclease. J Biol Chem. 2011; 286(14):12743±55. https://doi.org/ 10.1074/jbc.M110.199737 PMID: 21303903 64. Cross BC, Bond PJ, Sadowski PG, Jha BK, Zak J, Goodman JM, et al. The molecular basis for selective inhibition of unconventional mRNA splicing by an IRE1-binding small molecule. Proc Natl Acad Sci U S A. 2012; 109(15):E869±78. https://doi.org/10.1073/pnas.1115623109 PMID: 22315414 65. Mimura N, Fulciniti M, Gorgun G, Tai YT, Cirstea D, Santo L, et al. Blockade of XBP1 splicing by inhibition of IRE1alpha is a promising therapeutic option in multiple myeloma. Blood. 2012; 119(24):5772±81. https://doi.org/10.1182/blood-2011-07-366633 PMID: 22538852 66. Wiseman RL, Zhang Y, Lee KP, Harding HP, Haynes CM, Price J, et al. Flavonol activation defines an unanticipated ligand-binding site in the kinase-RNase domain of IRE1. Mol Cell. 2010; 38(2):291±304. https://doi.org/10.1016/j.molcel.2010.04.001 PMID: 20417606 67. Schindelin J, Arganda-Carreras I, Frise E, Kaynig V, Longair M, Pietzsch T, et al. Fiji: an open-source platform for biological-image analysis. Nature methods. 2012; 9(7):676±82. https://doi.org/10.1038/ nmeth.2019 PMID: 22743772 68. Rajpar MH, McDermott B, Kung L, Eardley R, Knowles L, Heeran M, et al. Targeted induction of endoplasmic reticulum stress induces cartilage pathology. PLoS Genet. 2009; 5(10):e1000691. https://doi. org/10.1371/journal.pgen.1000691 PMID: 19834559 69. Gal I, Varga T, Szilagyi I, Balazs M, Schlammadinger J, Szabo G, Jr. Protease-elicited TUNEL positivity of non-apoptotic fixed cells. J Histochem Cytochem. 2000; 48(7):963±70. https://doi.org/10.1177/ 002215540004800709 PMID: 10858273 70. Huang da W, Sherman BT, Lempicki RA. Bioinformatics enrichment tools: paths toward the compre- hensive functional analysis of large gene lists. Nucleic acids research. 2009; 37(1):1±13. https://doi.org/ 10.1093/nar/gkn923 PMID: 19033363 71. Huang da W, Sherman BT, Lempicki RA. Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources. Nat Protoc. 2009; 4(1):44±57. https://doi.org/10.1038/nprot.2008.211 PMID: 19131956 72. Draghici S, Khatri P, Tarca AL, Amin K, Done A, Voichita C, et al. A systems biology approach for pathway level analysis. Genome research. 2007; 17(10):1537±45. https://doi.org/10.1101/gr.6202607 PMID: 17785539

PLOS Genetics | https://doi.org/10.1371/journal.pgen.1008215 July 1, 2019 30 / 30